Underwater ultrasonic signal analysis method based on doppler effect

Abstract

The application provides a method for underwater ultrasonic signal analysis based on Doppler effect, and relates to the technical field of data processing.The method comprises the following steps: segmenting echo signal data according to a preset time window, performing envelope demodulation and phase unwrapping operation on the signal in each time window, and extracting amplitude sequence, phase sequence and instantaneous frequency sequence data; extracting frequency variation, amplitude variation and phase variation data, performing clustering processing, distinguishing and marking different underwater target components; calculating amplitude stability index and phase variation synchronization index for each frequency trajectory in all time windows, and screening main frequency trajectories according to the index; extracting frequency shift values in each time window, calculating the speed of each underwater target in each time window based on Doppler effect, and judging the moving direction in combination with frequency variation data and phase variation data; and the application improves the autonomy and accuracy of underwater ultrasonic signal analysis.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method for analyzing underwater ultrasonic signals based on the Doppler effect. Background Technology

[0002] Currently, existing technologies mainly employ the traditional Fourier transform method to perform spectral analysis on underwater ultrasonic signals. By analyzing the frequency shift of the received signal, the speed or distance of underwater targets can be estimated. This method typically combines underwater sonar equipment, first transmitting ultrasonic signals at a specific frequency, then receiving the echo signals reflected back from the underwater target. By calculating the frequency shift of the echo signal relative to the transmitted signal, the motion parameters of the underwater target can be inferred.

[0003] However, in practical applications, when the speed of underwater targets changes rapidly, or in environments with multiple underwater targets, traditional Fourier transform methods struggle to distinguish the Doppler frequency shift components of different underwater targets. For example, when monitoring multiple schools of fish simultaneously in a marine ranch, multiple reflected signals may overlap, leading to a decrease in frequency shift resolution. This can easily result in misidentification or loss of tracking of underwater targets, affecting the accuracy and reliability of underwater monitoring. Summary of the Invention

[0004] The purpose of this invention is to provide a method for analyzing underwater ultrasonic signals based on the Doppler effect, aiming to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] A method for analyzing underwater ultrasonic signals based on the Doppler effect, the method comprising:

[0007] The transmission frequency of the ultrasonic signal is obtained, and the ultrasonic signal is transmitted to the underwater target. The echo signals reflected back from multiple underwater targets are received, and echo signal data containing the reflection information of multiple underwater targets is obtained.

[0008] The echo signal data is segmented according to a preset time window. Envelope demodulation and phase expansion operations are performed on the signal within each time window to extract its amplitude sequence, phase sequence and instantaneous frequency sequence data, and generate corresponding time series signal feature data.

[0009] Based on the time series signal feature data, frequency change, amplitude change and phase change data are extracted, and clustering is performed to distinguish and label different underwater target components, resulting in a preliminary frequency shift signal set of multiple underwater targets. Each preliminary frequency shift signal set contains one or more frequency trajectories.

[0010] Based on the preliminary set of frequency shift signals, and combined with amplitude and phase change data, the amplitude stability index and phase change synchronization index are calculated for each frequency trajectory in all time windows, and the main frequency trajectory is selected accordingly to form the final set of frequency shift signals.

[0011] Based on the final frequency shift signal set, the frequency shift values ​​within each time window are extracted. The velocity of each underwater target within each time window is calculated based on the Doppler effect. The direction of motion is determined by combining the frequency change data and phase change data, and the motion state analysis results of the underwater targets are generated and output.

[0012] Preferably, the echo signal data is segmented according to a preset time window, and envelope demodulation and phase unrolling operations are performed on the signal within each time window to extract its amplitude sequence, phase sequence, and instantaneous frequency sequence data, generating corresponding time series signal feature data, including:

[0013] The echo signal data is framed along the time axis, and the echo signal is divided into multiple frame segments according to a preset time window to obtain a segmented signal sequence.

[0014] Based on the segmented signal sequence, envelope demodulation processing is performed on the echo signal of each frame segment to extract its amplitude variation curve and form amplitude sequence data.

[0015] Based on the echo signal data, a phase unrolling operation is performed to extract continuous phase change curves, and the phase growth rate of each frame segment is calculated to obtain phase sequence and instantaneous frequency sequence data.

[0016] The amplitude sequence, phase sequence, and instantaneous frequency sequence data are organized in chronological order to obtain the characteristic data of the time series signal.

[0017] Preferably, based on the time-series signal characteristic data, frequency change, amplitude change, and phase change data are extracted, clustering processing is performed, and different underwater target components are distinguished and labeled to obtain a preliminary set of frequency shift signals for multiple underwater targets, including:

[0018] Based on the characteristic data of the time series signal, the instantaneous frequency change, amplitude change and phase change data in each time window are extracted sequentially to form frequency change sequence, amplitude change sequence and phase change sequence respectively;

[0019] Based on the frequency change sequence, the frequency change amount of adjacent time windows is analyzed, and points with similar change trends are grouped into preliminary frequency change groups to obtain multiple frequency change candidate groups.

[0020] For each frequency change candidate group, its amplitude change sequence is extracted, and its mean and standard deviation are calculated respectively. Based on the condition that the mean amplitude is higher than the preset effective signal judgment threshold and the standard deviation is lower than the preset amplitude stability judgment threshold, the frequency change candidate groups that meet the amplitude criteria are selected.

[0021] For each candidate group of frequency changes that meets the criteria, analyze its phase change sequence, calculate the correlation between the phase change trend and the frequency change trend within the group, and determine the group with a correlation higher than the preset correlation threshold as the preliminary frequency shift signal set of the same underwater target. Each set contains one or more frequency trajectories.

[0022] Preferably, based on the preliminary frequency shift signal set, and combined with amplitude and phase change data, the amplitude stability index and phase change synchronization index are calculated for each frequency trajectory within all time windows, and the main frequency trajectory is selected accordingly to form the final frequency shift signal set, including:

[0023] For each frequency trajectory, extract its amplitude variation data over all time windows and calculate its standard deviation as an amplitude stability index.

[0024] For each frequency trajectory, based on the frequency change data and phase change data within all time windows, the frequency change direction and phase change direction within each time window are compared to determine whether the signs are consistent. The percentage of windows with consistent signs within all time windows is counted to obtain the consistency score of the growth direction.

[0025] For each frequency trajectory, the frequency change data and phase change data are first-order differencing to obtain the frequency change rate sequence and the phase change rate sequence. Correlation analysis is performed on the two at the same window position, and the proportion of windows with correlation higher than the preset synchronization judgment threshold is used as the phase change synchronization index.

[0026] For all frequency trajectories, combining the amplitude stability index, the growth direction consistency score, and the phase change synchronization index, a weighted scoring or threshold screening method is used to select the frequency trajectory with the highest comprehensive score as the main frequency trajectory. All main frequency trajectories are then collected to form the final frequency shift signal set.

[0027] Preferably, based on the final frequency shift signal set, the frequency shift values ​​within each time window are extracted, the velocity of each underwater target within each time window is calculated based on the Doppler effect, and the motion direction is determined by combining frequency change data and phase change data. The motion state analysis results of the underwater targets are then generated and output, including:

[0028] Based on the main frequency trajectory, the instantaneous frequency values ​​within each time window are extracted in the order of the time windows. Using the transmission frequency as a reference, the instantaneous frequency values ​​within each time window are differentially compared with the transmission frequency to obtain the frequency shift value of each time window.

[0029] Based on the frequency shift values ​​within each time window of each main frequency trajectory, the Doppler effect calculation method is used to divide the frequency shift values ​​by the transmission frequency and multiply them by the propagation speed of ultrasound in water to obtain the corresponding radial velocity sequence within each time window.

[0030] For each main frequency trajectory, based on the frequency change data and phase change data within each time window, determine the sign of the frequency change direction and the phase change direction within the same window. If they are consistent, determine that the underwater target is close to the source; if they are opposite, determine that the underwater target is far from the source. Generate motion direction criteria for each time window.

[0031] The radial velocity sequence of each main frequency trajectory within all time windows is combined with the motion direction criterion to form the velocity and motion direction data of each underwater target over the entire time series.

[0032] Based on the velocity and direction of motion data of each underwater target over the entire time series, the motion state analysis results of the underwater targets are generated and output.

[0033] Preferably, based on the segmented signal sequence, envelope demodulation processing is performed on the echo signal of each frame segment to extract its amplitude variation curve, forming amplitude sequence data, including:

[0034] For each frame of the segmented signal sequence, for each sampling point, the absolute value of the signal is calculated with that of multiple adjacent sampling points, and the average of the calculated absolute values ​​of the signal is performed in the time direction to obtain the envelope amplitude corresponding to that sampling point.

[0035] The envelope amplitudes of all sampling points in each frame are arranged sequentially according to the sampling time to form an amplitude variation curve.

[0036] Based on the amplitude variation curve, the amplitude at each sampling time is extracted point by point to form amplitude sequence data.

[0037] Preferably, based on the echo signal data, a phase unrolling operation is performed to extract continuous phase change curves, and the phase growth rate of each frame segment is calculated to obtain phase sequence data and instantaneous frequency sequence data, including:

[0038] Based on the echo signal data, for each sampling point, the preliminary phase value of the sampling point is calculated using its sine and cosine components, resulting in a set of preliminary phase sequence data;

[0039] The initial phase sequence data is subjected to continuity correction. When the phase difference between adjacent sampling points is found to be greater than half a cycle, the abrupt change is eliminated by adding or subtracting 2π, so that the phase change curve remains monotonically continuous in time, forming continuous phase sequence data.

[0040] Based on continuous phase sequence data, calculate the phase difference between every two adjacent sampling points, and divide the phase difference by the time interval between the two sampling points to obtain the phase growth rate sequence of each sampling point;

[0041] Arrange the phase growth rates of all sampling points in chronological order of sampling time to form instantaneous frequency sequence data.

[0042] Preferably, based on the frequency change sequence, the frequency changes in adjacent time windows are analyzed, and points with similar trends are grouped into preliminary frequency change groups to obtain multiple candidate frequency change groups, including:

[0043] Based on the frequency change sequence, the frequency change within each two adjacent time windows is differentially calculated to obtain the frequency change increment corresponding to each time window.

[0044] For the frequency change increments of all time windows, according to the principle of numerical similarity and consistent trend, the frequency change increments are clustered using the similarity criterion to obtain multiple frequency change subsequences with similar trends.

[0045] Connect the frequency changes of adjacent time windows in each frequency change subsequence sequentially to form preliminary frequency change groups;

[0046] Based on all the initial frequency change groupings, multiple frequency change candidate groups are generated.

[0047] Preferably, for each candidate group of frequency changes that meets the criteria, its phase change sequence is analyzed, the correlation between the phase change trend and the frequency change trend within the group is calculated, and groups with a correlation higher than a preset correlation threshold are identified as the preliminary frequency shift signal set for the same underwater target, including:

[0048] For each candidate group of frequency changes that meets the criteria, extract the phase change sequence within its corresponding time window to form the phase change data of that group.

[0049] For the frequency change sequence and phase change sequence of the candidate frequency change group, the first-order difference method is used to obtain the frequency change rate sequence and phase change rate sequence, respectively.

[0050] Correlation analysis was performed on the frequency change rate sequence and the phase change rate sequence within each corresponding time window, and the correlation coefficient was calculated to obtain the correlation criterion for each group.

[0051] For each frequency change candidate group, determine whether its correlation criterion is higher than the preset correlation threshold. Groups that are higher than the threshold are determined as the preliminary frequency shift signal set of the same underwater target, and output the preliminary frequency shift signal set that meets the conditions.

[0052] Preferably, for each frequency trajectory, its frequency change data and phase change data are first-order differencing to obtain a frequency change rate sequence and a phase change rate sequence, and correlation analysis is performed on the two at the same window position, including:

[0053] For each frequency trajectory, extract its frequency change data and phase change data within all time windows, and perform first-order difference processing according to the time window to obtain the corresponding frequency change rate sequence and phase change rate sequence.

[0054] Based on the frequency change rate sequence and the phase change rate sequence, at each time window position, the rate values ​​of the two are extracted respectively to construct a set of rate pairs under the same time window.

[0055] For rate pairs across all time windows, correlation analysis is performed using the correlation coefficient algorithm to obtain the correlation score sequence for each time window.

[0056] The above-described solution of the present invention has at least the following beneficial effects:

[0057] First, this invention is no longer limited to the single spectral analysis perspective of traditional Fourier transform. Instead, by segmenting the echo signal into time windows and employing envelope demodulation and phase unrolling operations, it can simultaneously extract the amplitude sequence, phase sequence, and instantaneous frequency sequence of the signal within each time window. This parallel extraction of multiple features greatly enriches the dimensions for capturing the physical properties of the echo signal, reflecting not only the target's velocity information but also real-time mapping of energy fluctuations and phase change trends during the target's motion.

[0058] Secondly, this invention utilizes multi-dimensional feature data (frequency changes, amplitude changes, and phase changes) for clustering processing, effectively distinguishing and labeling the signal components of different underwater targets in complex multi-target scenarios. Compared to the problem of overlapping and inaccurate separation of multi-target signals in existing technologies, this invention's solution, through multi-feature clustering, significantly improves the resolution of frequency-shift components and target separation capabilities, thereby reducing target misidentification and tracking loss, and enhancing the anti-interference capability and target detection accuracy of the underwater monitoring system.

[0059] Furthermore, when processing the initial frequency-shifted signal set, this invention not only considers the continuity of frequency changes but also integrates the stability of amplitude changes and the synchronicity of phase changes. It can automatically filter out the true, physically significant main frequency trajectories and further eliminate abnormal trajectories caused by noise, clutter, or multipath effects. For example, in a marine ranch monitoring scenario with dense fish schools and intersecting movement paths, the system can automatically lock onto the main movement trajectory of the real fish school, effectively avoiding interference from clutter and false signals.

[0060] Finally, by calculating the velocity of the main frequency trajectory based on the Doppler effect and combining it with frequency and phase change data to determine the target's motion direction, continuous and complete motion state analysis results can be output for each underwater target. This not only enhances the dynamic tracking capability of multiple underwater targets but also provides a solid data foundation for subsequent intelligent processing such as underwater target classification, behavior analysis, and path prediction.

[0061] In summary, this invention can significantly improve signal resolution and motion analysis accuracy in complex underwater environments with multiple targets. It solves the problems of poor frequency shift resolution, high false recognition rate, and tracking loss in existing technologies under scenarios with multiple targets, rapid movement, and noise interference. It has higher engineering application value and wider application prospects. Attached Figure Description

[0062] Figure 1 This is a flowchart of an underwater ultrasonic signal analysis method based on the Doppler effect provided in an embodiment of the present invention. Detailed Implementation

[0063] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0064] like Figure 1 As shown, embodiments of the present invention propose an underwater ultrasonic signal analysis method based on the Doppler effect, the method comprising:

[0065] The transmission frequency of the ultrasonic signal is obtained, and the ultrasonic signal is transmitted to the underwater target. The echo signals reflected back from multiple underwater targets are received, and echo signal data containing the reflection information of multiple underwater targets is obtained.

[0066] The echo signal data is segmented according to a preset time window. Envelope demodulation and phase expansion operations are performed on the signal within each time window to extract its amplitude sequence, phase sequence and instantaneous frequency sequence data, and generate corresponding time series signal feature data.

[0067] Based on the time series signal feature data, frequency change, amplitude change and phase change data are extracted, and clustering is performed to distinguish and label different underwater target components, resulting in a preliminary frequency shift signal set of multiple underwater targets. Each preliminary frequency shift signal set contains one or more frequency trajectories.

[0068] Based on the preliminary set of frequency shift signals, and combined with amplitude and phase change data, the amplitude stability index and phase change synchronization index are calculated for each frequency trajectory in all time windows, and the main frequency trajectory is selected accordingly to form the final set of frequency shift signals.

[0069] Based on the final frequency shift signal set, the frequency shift values ​​within each time window are extracted. The velocity of each underwater target within each time window is calculated based on the Doppler effect. The direction of motion is determined by combining the frequency change data and phase change data, and the motion state analysis results of the underwater targets are generated and output.

[0070] In this embodiment of the invention, by acquiring the transmission frequency of the emitted ultrasonic signal, transmitting it to an underwater target, and collecting the echo signals reflected back from multiple targets, a raw data foundation covering the diversity and dynamism of the targets can be obtained. The echo signal data is processed in segments according to a preset time window, and through envelope demodulation and phase unfolding operations, amplitude sequence, phase sequence, and instantaneous frequency sequence data are extracted respectively, achieving comprehensive quantization of the multidimensional physical characteristics of the original signal.

[0071] In subsequent processing, cluster analysis based on the extracted three types of change data (frequency, amplitude, and phase) can effectively distinguish the echo components of multiple underwater targets. Even in complex environments (such as multiple targets, overlapping, and noisy fields), the signal trajectories of different targets can be separated to obtain a preliminary set of frequency-shifted signals.

[0072] Further filtering based on amplitude stability and phase synchronization can automatically eliminate abnormal trajectories, retaining only the physically dominant frequency trajectory, greatly improving the accuracy and robustness of the dominant trajectory. For example, if a frequency trajectory has large amplitude fluctuations and irregular phase changes, it is automatically identified as interference; conversely, a stable trajectory can be accurately marked as the target motion path.

[0073] Ultimately, by extracting frequency shift data from the main frequency trajectory, calculating the velocity of each target based on the Doppler effect, and combining frequency and phase change data to determine the direction of motion, the system achieved full dynamic tracking of velocity and direction and high-resolution motion state analysis of all underwater targets.

[0074] For example:

[0075] In real-world underwater sonar scenarios, if two moving vessels and a background obstacle exist simultaneously, the method of this invention can separate and label their respective reflected signals, calculate the speed and approach or departure status of each vessel, and eliminate the influence of background noise, providing complete and reliable analysis results for engineering applications such as underwater target identification, navigation, and obstacle avoidance.

[0076] Through the above series of processing steps, this invention significantly improves the accuracy of signal recognition, the continuity of motion trajectory tracking, and the practicality of terminal applications in multi-target dynamic underwater environments.

[0077] In a preferred embodiment of the present invention, the echo signal data is segmented according to a preset time window, and envelope demodulation and phase unrolling operations are performed on the signal within each time window to extract its amplitude sequence, phase sequence, and instantaneous frequency sequence data, generating corresponding time-series signal feature data, including:

[0078] The echo signal data is framed along the time axis, and the echo signal is divided into multiple frame segments according to a preset time window to obtain a segmented signal sequence.

[0079] Based on the segmented signal sequence, envelope demodulation processing is performed on the echo signal of each frame segment to extract its amplitude variation curve and form amplitude sequence data.

[0080] Based on the echo signal data, a phase unrolling operation is performed to extract continuous phase change curves, and the phase growth rate of each frame segment is calculated to obtain phase sequence and instantaneous frequency sequence data.

[0081] The amplitude sequence, phase sequence, and instantaneous frequency sequence data are organized in chronological order to obtain the characteristic data of the time series signal.

[0082] In this embodiment of the invention, by dividing the echo signal data into time-axis frames, local changes in the echo signal can be captured with higher temporal resolution. For example, in ocean observation, if a target suddenly accelerates or decelerates, frame division can record this process in detail, avoiding the loss of local dynamic features due to overall averaging.

[0083] For each frame segment, envelope demodulation is employed to effectively eliminate the influence of high-frequency carrier waves, directly obtaining amplitude variation curves reflecting the distance and volume of the target object. Subsequently, phase unrolling is performed on the original echo signal to obtain continuous phase change information, allowing for the calculation of the phase growth rate and instantaneous frequency for each frame segment. For example, the phase and frequency data of the same target across different frame segments can clearly reflect minute changes in its direction and velocity.

[0084] Finally, by organizing the amplitude sequence, phase sequence, and instantaneous frequency sequence in chronological order, complete time-series signal feature data is generated. This not only supports continuous analysis of multiple targets and events but also provides a stable, rich, and easily processed source of raw data for subsequent multidimensional feature clustering and fine separation of underwater targets. For example, in complex port environments, it can distinguish the different echo characteristics of berthed ships and distant vessels.

[0085] Specifically, the amplitude sequence, phase sequence, and instantaneous frequency sequence data are organized in chronological order to obtain time-series signal characteristic data, which includes:

[0086] First, after processing the echo signal of each frame segment, amplitude sequence, phase sequence, and instantaneous frequency sequence are obtained, arranged according to the sampling time. Each set of data is ordered according to the start and end sampling points of the time window, ensuring that each feature value corresponds to the sampling point at the same time.

[0087] Next, these three sets of data are synchronized and paired along the "time" axis. That is, at each sampling moment, there is a corresponding amplitude, a phase, and an instantaneous frequency characteristic value. In this way, all feature points are arranged in chronological order.

[0088] Finally, the amplitude, phase, and instantaneous frequency arranged in time sequence under each time window are combined to form a complete set of time-labeled multidimensional signal feature data, which is used for subsequent target identification and cluster analysis.

[0089] For example, assuming the time window is 10 milliseconds and each sampling point is spaced 1 millisecond apart, then 10 sets of amplitude, phase and instantaneous frequency feature points will be arranged sequentially within each window, and the corresponding sampling time point will be clearly marked in each set of data.

[0090] In a preferred embodiment of the present invention, frequency change, amplitude change, and phase change data are extracted based on time series signal feature data, and clustering processing is performed to distinguish and label different underwater target components, resulting in a preliminary set of frequency shift signals for multiple underwater targets, including:

[0091] Based on the characteristic data of the time series signal, the instantaneous frequency change, amplitude change and phase change data in each time window are extracted sequentially to form frequency change sequence, amplitude change sequence and phase change sequence respectively;

[0092] Based on the frequency change sequence, the frequency change amount of adjacent time windows is analyzed, and points with similar change trends are grouped into preliminary frequency change groups to obtain multiple frequency change candidate groups.

[0093] For each frequency change candidate group, its amplitude change sequence is extracted, and its mean and standard deviation are calculated respectively. Based on the condition that the mean amplitude is higher than the preset effective signal judgment threshold and the standard deviation is lower than the preset amplitude stability judgment threshold, the frequency change candidate groups that meet the amplitude criteria are selected.

[0094] For each candidate group of frequency changes that meets the criteria, analyze its phase change sequence, calculate the correlation between the phase change trend and the frequency change trend within the group, and determine the group with a correlation higher than the preset correlation threshold as the preliminary frequency shift signal set of the same underwater target. Each set contains one or more frequency trajectories.

[0095] In this embodiment of the invention, the instantaneous frequency change, amplitude change, and phase change within each time window are sequentially extracted from the time series signal feature data, which can decompose the original complex signal into a multi-dimensional data sequence that is easy to analyze. Taking a practical application as an example, if a target (such as a submarine) passes through the monitoring area at a constant speed, its frequency change trend, amplitude stability, and phase synchronization can be accurately tracked, thereby distinguishing it from other small clutter or interference targets.

[0096] Subsequently, frequency variation trend analysis was employed to group data points with similar trends into preliminary frequency variation groups, which helps to initially identify target segments with consistent motion states. Further, for each candidate frequency variation group, the mean and standard deviation of the amplitude variation sequence were calculated to filter out signal groups with high amplitude and low fluctuations. For example, when large marine organisms or underwater robots move, their echo amplitudes are often large and stable, effectively eliminating low-amplitude noise.

[0097] Finally, for candidate groups that meet the amplitude criteria, their phase change sequences are further analyzed, and correlation analysis is performed with the frequency change trend. Only those groups whose phase and frequency change synchronously are retained as valid targets. For example, when an underwater target approaches the sonar array in a straight line, its frequency and phase change trends are highly synchronized, allowing for accurate identification and tracking. Through this series of progressively refined and anomaly-eliminating processes, the final preliminary set of frequency-shift signals will only contain valid underwater target trajectories with obvious physical characteristics and clear motion states, providing high-quality input for subsequent high-resolution estimation of velocity and direction.

[0098] Specifically, based on the characteristic data of the time series signal, the instantaneous frequency change, amplitude change, and phase change data within each time window are extracted sequentially to form frequency change sequences, amplitude change sequences, and phase change sequences, respectively, including:

[0099] First, the amplitude, phase, and instantaneous frequency feature points of the same target are extracted from the time series signal feature data, and arranged into their respective initial sequences according to time sequence.

[0100] For a frequency change sequence, the instantaneous frequency characteristic values ​​within two adjacent time windows are subtracted sequentially to obtain the frequency change within each window. These changes are then arranged sequentially to obtain the frequency change sequence.

[0101] Similarly, for the amplitude change sequence, the amplitude characteristic values ​​of adjacent time windows are subtracted to obtain the amplitude change amount, and then arranged in order to form the amplitude change sequence.

[0102] For a phase change sequence, the phase characteristic value changes between adjacent windows are calculated sequentially and arranged in order to obtain the phase change sequence.

[0103] This differential extraction method can accurately reflect the dynamic changes of the target within each time window, providing a rich data foundation for subsequent clustering, target identification, and abnormal trajectory screening.

[0104] For example, if the instantaneous frequency characteristics of three consecutive time windows are 10kHz, 10.5kHz, and 11kHz, then the frequency change sequence is 0.5kHz, 0.5kHz; similarly, the generation methods of amplitude and phase change sequences are exactly the same.

[0105] The methods for determining the preset effective signal determination threshold and the preset amplitude stability determination threshold specifically include:

[0106] Valid signal determination threshold:

[0107] Generally, the amplitude statistical characteristics of the echo signal are used for setting. Specifically, a set of representative underwater echo signals under background noise conditions can be selected, their amplitude distribution can be statistically analyzed, and a lower limit for the amplitude above the background noise can be set. For example, the average amplitude under background noise conditions plus a certain number of times the standard deviation can be used as the minimum amplitude criterion for determining that the echo signal is an "effective target reflection signal".

[0108] During implementation, historical signals can be measured to statistically analyze the maximum amplitude in multiple targetless environments. Based on this, a higher judgment threshold can be set so that only signals with amplitudes exceeding the threshold are considered valid target signals, thus filtering out environmental noise and clutter.

[0109] Amplitude stability judgment threshold:

[0110] It is mainly used to determine the stability of signal amplitude within a target motion cycle. The specific method is as follows: First, calculate the standard deviation of the amplitude sequence for each target candidate group; the smaller the standard deviation, the more stable the signal amplitude change. By collecting a large number of echo signals from different targets under different motion states, statistically analyze the standard deviation distribution of each amplitude sequence.

[0111] Based on practical application needs, a maximum permissible standard deviation threshold is preset. For example, in previous fish swarm monitoring tasks, if the amplitude standard deviation is greater than the set value, the group is very likely to be subject to sudden interference or signal obstruction, and cannot be used as a stable target trajectory. Usually, an empirical method (such as using the standard deviation of more than 95% of the target amplitude stability as a reference) is chosen, or adaptive optimization is performed using machine learning methods to continuously adjust the threshold to adapt to different environmental noise levels and signal quality requirements.

[0112] For example, if the amplitude standard deviation is 0.5 under background noise level, and the effective signal amplitude of the target is mostly higher than 1.0, in practical applications, the effective signal judgment threshold can be set to 1.2 and the amplitude stability judgment threshold can be set to 0.8 to ensure the accuracy and robustness of target recognition.

[0113] In a preferred embodiment of the present invention, based on the preliminary frequency shift signal set and combined with amplitude and phase change data, the amplitude stability index and phase change synchronization index are calculated for each frequency trajectory within all time windows, and the main frequency trajectory is selected accordingly to form the final frequency shift signal set, including:

[0114] For each frequency trajectory, extract its amplitude variation data over all time windows and calculate its standard deviation as an amplitude stability index.

[0115] For each frequency trajectory, based on the frequency change data and phase change data within all time windows, the frequency change direction and phase change direction within each time window are compared to determine whether the signs are consistent. The percentage of windows with consistent signs within all time windows is counted to obtain the consistency score of the growth direction.

[0116] For each frequency trajectory, the frequency change data and phase change data are first-order differencing to obtain the frequency change rate sequence and the phase change rate sequence. Correlation analysis is performed on the two at the same window position, and the proportion of windows with correlation higher than the preset synchronization judgment threshold is used as the phase change synchronization index.

[0117] For all frequency trajectories, combining the amplitude stability index, the growth direction consistency score, and the phase change synchronization index, a weighted scoring or threshold screening method is used to select the frequency trajectory with the highest comprehensive score as the main frequency trajectory. All main frequency trajectories are then collected to form the final frequency shift signal set.

[0118] In this embodiment of the invention, by calculating the amplitude stability index and phase change synchronization index for each frequency trajectory in the preliminary frequency shift signal set within all time windows, the main frequency trajectory with the strongest physical meaning and the best signal quality can be selectively selected.

[0119] For example, suppose underwater sonar detects three reflecting targets (e.g., A, B, and C). The initial frequency trajectory of each target may fluctuate due to noise or obstruction. By statistically analyzing the standard deviation of the amplitude variation data for each trajectory, interference trajectories with large amplitude fluctuations and poor stability can be eliminated. For instance, if target B experiences sudden amplitude jitter due to local bubble interference, its standard deviation will be significantly higher than that of A and C, and it will be judged as a non-dominant trajectory.

[0120] Furthermore, a synchronicity analysis is performed on the phase change data of the frequency trajectory, calculating the consistency between the direction of frequency change and the direction of phase change within each time window, as well as the correlation of rate changes. This ensures that the selected main frequency trajectory maintains consistent frequency and phase change trends throughout the entire time dimension. For example, if target A approaches at a constant linear speed while target C accelerates away, both exhibit a high degree of synchronicity in their frequency and phase changes. Therefore, the synchronicity index ensures that the main frequency trajectory reflects the true target motion.

[0121] This technology can greatly improve the physical accuracy and anti-interference capability of the final frequency shift signal set, effectively avoid the introduction of false trajectories, and provide a strong data foundation for subsequent high-resolution analysis of velocity and direction.

[0122] Specifically, for all frequency trajectories, combining amplitude stability index, growth direction consistency score, and phase change synchronization index, a weighted scoring or threshold screening method is used to select the frequency trajectory with the highest comprehensive score as the main frequency trajectory. All main frequency trajectories are then collected to form the final frequency shift signal set, which specifically includes:

[0123] Feature extraction:

[0124] For each frequency trajectory, firstly, the amplitude variation data within all time windows are extracted, and the standard deviation of the amplitude variation is calculated as the "amplitude stability index" for that trajectory. Secondly, the sign consistency between the frequency variation direction and the phase variation direction within each time window is statistically analyzed, and the proportion of windows with consistent signs is obtained as the "growth direction consistency score". Thirdly, the frequency variation data and phase variation data are respectively subjected to first-order difference to obtain their respective rate sequences. Then, the correlation analysis of the rates in the corresponding time windows is performed, and the proportion of windows with correlation higher than a set threshold is statistically analyzed as the "phase variation synchronization index".

[0125] Feature normalization:

[0126] The three indicators mentioned above are normalized (e.g., uniformly adjusted to a range between 0 and 1) to eliminate the influence of different dimensions and ranges on the subsequent weighted scoring.

[0127] Weighted score calculation:

[0128] Set weight coefficients for the three indicators of amplitude stability, consistency of growth direction, and synchronicity of phase change (e.g., 0.4, 0.3, and 0.3 respectively, which can be adjusted based on historical data, experimental results, or actual scenarios). Then, multiply the normalized score of each indicator by the corresponding weight coefficient, and add the three results together to obtain the "comprehensive score" of the frequency trajectory.

[0129] Example description: If the normalized scores of trajectory A are 0.8, 0.7, and 0.9 respectively, then the final weighted score = 0.8 × 0.4 + 0.7 × 0.3 + 0.9 × 0.3.

[0130] Main trajectory filtering and aggregation:

[0131] After calculating weighted scores for all frequency trajectories, the trajectory with the highest overall score within each initial frequency shift signal set is selected as the master frequency trajectory. If threshold screening is used, a minimum score threshold can be set, and only trajectories with scores higher than the threshold are retained as master trajectories.

[0132] Forming the final set:

[0133] The main frequency trajectories selected from all the preliminary frequency shift signal sets are collected and output as the "final frequency shift signal set" for subsequent motion parameter analysis and output.

[0134] In a preferred embodiment of the present invention, based on the final frequency shift signal set, the frequency shift values ​​within each time window are extracted, the velocity of each underwater target within each time window is calculated based on the Doppler effect, and the motion direction is determined by combining frequency change data and phase change data, generating and outputting the motion state analysis results of the underwater targets, including:

[0135] Based on the main frequency trajectory, the instantaneous frequency values ​​within each time window are extracted in the order of the time windows. Using the transmission frequency as a reference, the instantaneous frequency values ​​within each time window are differentially compared with the transmission frequency to obtain the frequency shift value of each time window.

[0136] Based on the frequency shift values ​​within each time window of each main frequency trajectory, the Doppler effect calculation method is used to divide the frequency shift values ​​by the transmission frequency and multiply them by the propagation speed of ultrasound in water to obtain the corresponding radial velocity sequence within each time window.

[0137] For each main frequency trajectory, based on the frequency change data and phase change data within each time window, determine the sign of the frequency change direction and the phase change direction within the same window. If they are consistent, determine that the underwater target is close to the source; if they are opposite, determine that the underwater target is far from the source. Generate motion direction criteria for each time window.

[0138] The radial velocity sequence of each main frequency trajectory within all time windows is combined with the motion direction criterion to form the velocity and motion direction data of each underwater target over the entire time series.

[0139] Based on the velocity and direction of motion data of each underwater target over the entire time series, the motion state analysis results of the underwater targets are generated and output.

[0140] In this embodiment of the invention, the frequency shift values ​​within each time window are extracted using the final frequency shift signal set, and the velocity is calculated based on the Doppler effect, enabling precise quantification of the radial velocity of each underwater target at different time periods. For example, assuming target A is actively approaching and target B is slowly moving away, the instantaneous frequency is extracted in each time window using their respective main frequency trajectories. After subtracting from the transmission frequency, the corresponding velocity sequence is calculated using the Doppler formula. In actual testing, even slight changes in target velocity can be reflected in real time, dynamically, and accurately.

[0141] Furthermore, by combining the frequency and phase change data of each main frequency trajectory, it is determined whether the target is approaching or moving away within each window (sign judgment), thereby obtaining directional velocity information. For example, if both frequency and phase increase within a certain window, it indicates that the target is rapidly approaching; conversely, it indicates that the target is moving away. Finally, by combining the velocity and motion direction criteria of each target across all windows, a complete dynamic state curve of the underwater target over time can be output.

[0142] This step significantly improves the accuracy and real-time performance of target motion state estimation, meeting the needs of continuous monitoring, situational awareness, and behavior prediction in multi-target, highly dynamic scenarios. For example, in complex fishing grounds or ports, it can distinguish the motion trajectories and speed characteristics of fishing boats, submersibles, and naturally floating objects, assisting in automated early warning and intelligent underwater management.

[0143] Specifically, the radial velocity sequence of each main frequency trajectory within all time windows is combined with the motion direction criterion to form the velocity and motion direction data of each underwater target over the entire time series, including:

[0144] Extraction of the main frequency trajectory velocity sequence:

[0145] For each main frequency trajectory, the instantaneous frequency values ​​within each window are extracted sequentially according to the time window order. The frequency shift value is obtained by subtracting the transmission frequency. The ratio of the frequency shift value to the transmission frequency is then multiplied by the ultrasonic wave propagation speed in water to obtain the radial velocity data within each window.

[0146] For example, if the transmission frequency is 50kHz and the instantaneous frequency of the main trajectory is 51kHz in a certain time window, then the frequency shift value is 1kHz and the radial velocity is (1kHz / 50kHz)×1500m / s=30m / s.

[0147] Criteria for determining the direction of motion:

[0148] For each time window of the main frequency trajectory, analyze the sign relationship between the direction of frequency change and the direction of phase change within that window:

[0149] When the signs are consistent (e.g., both are positive or both are negative), the underwater target's movement direction within that time window is determined to be "approaching" the source.

[0150] When the signs are opposite, it is determined that the source is "far away".

[0151] This criterion is coded as "1" or "-1", or simply recorded as near / far from the label.

[0152] Combination of velocity and direction data:

[0153] The radial velocity sequence of the main frequency trajectory is combined with the corresponding motion direction criteria to form a complete "velocity + direction" feature point at each moment (time window).

[0154] Generation of target motion states throughout the entire sequence:

[0155] The velocity and direction of motion data for all time windows under each main trajectory are combined in chronological order to form the complete motion state time series data of the underwater target within the monitoring period. If there are multiple main frequency trajectories, the velocity and direction of motion time series data of the corresponding underwater target are generated separately.

[0156] Output results:

[0157] The final output includes the radial velocity and direction of motion of each underwater target (main frequency trajectory) in each time window, which is convenient for subsequent applications such as target tracking, classification, risk warning or underwater behavior analysis.

[0158] In a preferred embodiment of the present invention, envelope demodulation processing is performed on the echo signal of each frame segment according to the segmented signal sequence to extract its amplitude variation curve and form amplitude sequence data, including:

[0159] For each frame of the segmented signal sequence, for each sampling point, the absolute value of the signal is calculated with that of multiple adjacent sampling points, and the average of the calculated absolute values ​​of the signal is performed in the time direction to obtain the envelope amplitude corresponding to that sampling point.

[0160] The envelope amplitudes of all sampling points in each frame are arranged sequentially according to the sampling time to form an amplitude variation curve.

[0161] Based on the amplitude variation curve, the amplitude at each sampling time is extracted point by point to form amplitude sequence data.

[0162] In this embodiment of the invention, by performing envelope demodulation processing on each frame of echo signal through segmented signal sequences, the masking of amplitude characteristics by the high-frequency carrier can be effectively eliminated, and the amplitude variation curve that truly reflects the target volume, distance, and signal propagation loss can be extracted. For example, in an underwater environment, if the volume of a reflective target (such as a school of fish) suddenly changes or the signal propagation path is disturbed, the envelope curve will respond quickly and sensitively, showing local peaks or troughs in the amplitude.

[0163] Furthermore, by calculating the absolute values ​​of the signals at all sampling points within each frame segment and their adjacent sampling points, and then performing a time-averaging operation, local high-frequency noise can be effectively smoothed and the continuity of the amplitude curve can be improved. For example, if there are weak background clutter or high-frequency mechanical interference underwater, their impact on the amplitude sequence is significantly reduced after the above averaging process.

[0164] Finally, by extracting the amplitude values ​​at each sampling moment point by point and arranging them in chronological order, a complete and detailed amplitude sequence data can be formed. This data can not only be used for subsequent cluster analysis and target identification, but also facilitates a visual representation of the signal intensity evolution of each target in sonar monitoring, providing intuitive and continuous data support for real-time monitoring and historical analysis. For example, in response to sudden underwater events (such as explosions or large-scale fish migrations), rapid and abnormal changes in the amplitude sequence can enable timely detection and response.

[0165] Specifically, for each frame segment in the segmented signal sequence, for each sampling point, the absolute value of the signal between the sampling point and its neighboring sampling points is calculated, and the calculated absolute values ​​are averaged over time to obtain the envelope amplitude corresponding to that sampling point. This includes:

[0166] Sampling point and neighborhood selection:

[0167] For each frame of signal, each sampling point is traversed sequentially from beginning to end. For the current sampling point, several adjacent sampling points (e.g., 2 to 5 each) are selected to form a local neighborhood sampling window. The size of the neighborhood window can be set according to the actual sampling rate and signal characteristics.

[0168] Absolute value statistical processing:

[0169] For the current sampling point and all sampling points in its neighborhood, take the absolute value of its amplitude (i.e. remove the positive and negative signs and retain only the signal strength value).

[0170] Time average calculation:

[0171] The absolute values ​​of all sampling points within the aforementioned neighborhood window are summed, and then averaged using the number of sampling points to obtain the envelope amplitude of the current sampling point.

[0172] For example, if the absolute value of a sampling point and its four neighboring sampling points is {3,4,5,6,7}, then its envelope amplitude is (3+4+5+6+7) / 5=5.

[0173] Full-frame envelope extraction:

[0174] The above steps are repeated for all sampling points within each frame segment to sequentially generate the envelope amplitude sequence of all sampling points in that frame segment, thus completing the envelope demodulation processing of one frame of signal. This operation can be expanded frame by frame across the entire segmented signal sequence, ultimately forming continuous amplitude variation curves and amplitude sequence data, providing an input basis for subsequent analysis.

[0175] In a preferred embodiment of the present invention, a phase unfolding operation is performed based on the echo signal data to extract continuous phase change curves, and the phase growth rate of each frame segment is calculated to obtain phase sequence data and instantaneous frequency sequence data, including:

[0176] Based on the echo signal data, for each sampling point, the preliminary phase value of the sampling point is calculated using its sine and cosine components, resulting in a set of preliminary phase sequence data;

[0177] The initial phase sequence data is subjected to continuity correction. When the phase difference between adjacent sampling points is found to be greater than half a cycle, the abrupt change is eliminated by adding or subtracting 2π, so that the phase change curve remains monotonically continuous in time, forming continuous phase sequence data.

[0178] Based on continuous phase sequence data, calculate the phase difference between every two adjacent sampling points, and divide the phase difference by the time interval between the two sampling points to obtain the phase growth rate sequence of each sampling point;

[0179] Arrange the phase growth rates of all sampling points in chronological order of sampling time to form instantaneous frequency sequence data.

[0180] In this embodiment of the invention, by performing a phase unfolding operation on the echo signal data, extracting continuous phase change curves, and calculating the phase growth rate of each frame segment, high-precision interpretation and quantification of the echo signal phase information can be achieved.

[0181] For example, for target object A, since its distance relative to the emission source is constantly changing, the phase of its reflected signal will change continuously. By calculating the preliminary phase from the sine and cosine components of each sampling point, and combining this with continuity correction (such as adding or subtracting 2π to eliminate abrupt changes), problems such as phase jumps and unwrapping discontinuities can be effectively avoided, ensuring that the phase curve within each time window truly and consistently reflects the target's motion process.

[0182] Furthermore, the phase difference between adjacent sampling points is calculated using continuous phase sequence data, and the phase growth rate is calculated by combining the time interval, thereby obtaining an instantaneous frequency sequence that reflects the minute dynamic changes of the signal.

[0183] For example, when a submersible moving at high speed passes by, even if there are slight changes in speed and trajectory, its instantaneous frequency curve can be reflected synchronously, achieving high-resolution dynamic tracking.

[0184] The above methods greatly improve the tracking accuracy and robustness of complex underwater moving targets, providing a solid signal foundation for subsequent velocity and motion direction discrimination based on the Doppler effect.

[0185] Specifically, based on the echo signal data, for each sampling point, the preliminary phase value of that sampling point is calculated using its sine and cosine components, resulting in a set of preliminary phase sequence data, which includes:

[0186] Raw data preparation:

[0187] For the original echo signal within each frame segment, each sampling point is traversed sequentially, and the amplitude of that sampling point in the original signal is extracted respectively.

[0188] Obtaining sine and cosine components:

[0189] For each sampling point, based on common principles of digital signal processing, the sine component (real part) and cosine component (imaginary part) of that sampling point are calculated. In practical applications, the Hilbert transform of the original signal can be used to obtain its analytic signal, thereby directly extracting the real and imaginary parts of each sampling point.

[0190] Preliminary phase value calculation:

[0191] For each sampling point, the cosine component is used as the abscissa and the sine component as the ordinate. Based on the arctangent relationship of trigonometric functions, the preliminary phase value of the sampling point is calculated using the arctangent calculation method (such as the arctan or atan2 function).

[0192] In layman's terms, this means that the real and imaginary parts of each sampling point on the two-dimensional plane are transformed into a unique phase angle through polar coordinate transformation.

[0193] Preliminary phase sequence generation:

[0194] The preliminary phase values ​​calculated from each sampling point are arranged in chronological order of sampling time to obtain the complete preliminary phase sequence data for that frame segment. This data sequence provides the foundation for subsequent phase expansion processing (removing 2π transitions and ensuring continuity) and further instantaneous frequency calculations.

[0195] In a preferred embodiment of the present invention, based on the frequency change sequence, the frequency change amount of adjacent time windows is analyzed, and points with similar change trends are grouped into preliminary frequency change groups to obtain multiple frequency change candidate groups, including:

[0196] Based on the frequency change sequence, the frequency change within each two adjacent time windows is differentially calculated to obtain the frequency change increment corresponding to each time window.

[0197] For the frequency change increments of all time windows, according to the principle of numerical similarity and consistent trend, the frequency change increments are clustered using the similarity criterion to obtain multiple frequency change subsequences with similar trends.

[0198] Connect the frequency changes of adjacent time windows in each frequency change subsequence sequentially to form preliminary frequency change groups;

[0199] Based on all the initial frequency change groupings, multiple frequency change candidate groups are generated.

[0200] In this embodiment of the invention, the frequency change amount of adjacent time windows is analyzed according to the frequency change sequence, and points with similar change trends are grouped into preliminary frequency change groups, which can realize intelligent screening and clustering of frequency components in multi-objective scenarios.

[0201] For example, when multiple moving targets exist within the sonar detection range, each target has different frequency variation characteristics in its reflected signal due to its different relative velocities. By clustering the frequency variation increments within all time windows using the principles of numerical proximity and consistent trends, the frequency components corresponding to multiple targets can be automatically grouped.

[0202] If targets A, B, and C are in a state of uniform approach, moving away, and accelerating approach respectively, clustering can distinguish the frequency change subsequences related to target A, the grouping of target B, and the grouping of target C, thus laying the foundation for subsequent detailed analysis and classification labeling.

[0203] This intelligent grouping effectively reduces data dimensionality, minimizes human intervention, and improves target recognition efficiency and automation, making it particularly suitable for complex underwater environments with high noise and interference.

[0204] Specifically, for the frequency change increments across all time windows, a similarity criterion is used to cluster the frequency change increments based on the principles of numerical similarity and consistent trends, resulting in multiple frequency change subsequences with similar trends, including:

[0205] Calculation and organization of frequency change increments:

[0206] First, for all time windows of frequency change data, calculate the frequency change increment for every two adjacent time windows, organize all increment results into a sequence, and number them in chronological order.

[0207] Similarity criterion design:

[0208] For the frequency change increment, two main similarity criteria are set:

[0209] The first is "numerical proximity," which means that if the difference between two frequency change increments is within a certain range (such as the threshold range set by empirical values), they are considered to be numerically close.

[0210] Second, "consistency of change trend" means that if the signs of the increments of two frequencies are the same (e.g., both are increasing or both are decreasing), then their change trends are considered to be consistent.

[0211] Clustering:

[0212] A stepwise traversal and recursive clustering method is adopted. Starting from the first time window, subsequent increments are sequentially judged to see if they simultaneously meet the two conditions of "numerical proximity" and "consistency of change trend". If they are met, these frequency change increments are merged into the same group to form a "frequency change subsequence". When an increment that does not meet the above conditions is encountered, a new grouping is initiated.

[0213] Output clustering results:

[0214] Using the above method, all frequency change increments are ultimately divided into several different frequency change subsequences with similar trends. Each subsequence corresponds to a target frequency change component with obvious consistency characteristics, laying the foundation for subsequent amplitude criteria and target clustering.

[0215] In a preferred embodiment of the present invention, for each candidate group of frequency changes that meets the frame criterion, its phase change sequence is analyzed, the correlation between the phase change trend and the frequency change trend within the group is calculated, and groups with a correlation higher than a preset correlation threshold are determined as a preliminary set of frequency shift signals for the same underwater target, including:

[0216] For each candidate group of frequency changes that meets the criteria, extract the phase change sequence within its corresponding time window to form the phase change data of that group.

[0217] For the frequency change sequence and phase change sequence of the candidate frequency change group, the first-order difference method is used to obtain the frequency change rate sequence and phase change rate sequence, respectively.

[0218] Correlation analysis was performed on the frequency change rate sequence and the phase change rate sequence within each corresponding time window, and the correlation coefficient was calculated to obtain the correlation criterion for each group.

[0219] For each frequency change candidate group, determine whether its correlation criterion is higher than the preset correlation threshold. Groups that are higher than the threshold are determined as the preliminary frequency shift signal set of the same underwater target, and output the preliminary frequency shift signal set that meets the conditions.

[0220] In this embodiment of the invention, for each candidate group of frequency changes that meets the criteria for image quality, the phase change sequence is analyzed, and the correlation between the phase change trend and the frequency change trend within the group is calculated. Groups with a correlation higher than a preset correlation threshold are identified as the initial frequency shift signal set of the same underwater target, which can further improve the accuracy of target separation and the physical reliability of clustering.

[0221] For example, among several candidate frequency change groups that have been screened for amplitude stability, some groups, although stable in amplitude, do not exhibit physical synchronicity between frequency and phase changes (e.g., abnormal groups caused by reflection noise or multipath aliasing). In this case, by performing correlation coefficient analysis on the frequency change rate sequence and phase change rate sequence within each group, physically irrelevant or noise-dominated groups are eliminated, retaining only those groups with high correlation and strong synchronicity as the initial set of target frequency shift signals.

[0222] For example, if target A and target B are in the same direction but have different speeds, and their phase and frequency changes are highly synchronized, they can be correctly classified as two independent and real targets, preventing missed detections or false detections.

[0223] This step can significantly improve the physical rationality and practical effectiveness of target differentiation and clustering in multi-target environments, providing a strong foundation for subsequent main frequency trajectory screening and motion state analysis.

[0224] This involves performing correlation analysis on the frequency change rate sequence and the phase change rate sequence within their respective time windows, calculating the correlation coefficient, and obtaining the correlation criteria for each group, specifically including:

[0225] Sequence preparation:

[0226] For each candidate group of frequency changes, first obtain its frequency change rate sequence and phase change rate sequence for each time window. The rate sequences here are all sequence data obtained from previous steps (such as first-order difference), and are arranged in order of time window.

[0227] Corresponding window pairing:

[0228] For the same time window, the frequency change rate and phase change rate of the group are matched one-to-one to form rate pairs under the same window.

[0229] Correlation coefficient analysis method:

[0230] For the above rate pairs, a sliding window approach is used to select a time range (such as 5 to 10 consecutive time windows) and count the degree of coordination of all rate pairs within this range.

[0231] In specific processing, standard correlation coefficient statistical methods (such as the Pearson correlation coefficient principle) are used, that is, comparing the degree of synchronization between the direction and amplitude of the frequency change rate and the phase change rate one by one.

[0232] If the trends of increase or decrease over time are consistent, the correlation coefficient is close to 1; if the trends are opposite, the correlation coefficient is negative or close to -1; if there is no obvious relationship, the correlation coefficient is close to 0.

[0233] Correlation criterion output:

[0234] For all time windows, the correlation score of each frequency-phase rate pair is calculated, and a correlation criterion sequence is obtained. Based on the statistical results, it is then determined which time windows have a correlation higher than a set judgment threshold to determine whether the group is an effective frequency shift signal set of the same underwater target.

[0235] The method for determining the preset correlation threshold specifically includes:

[0236] Historical data statistical analysis:

[0237] By analyzing a large number of historical underwater ultrasonic signal samples, the frequency change rate and phase change rate sequences of the same known target were selected, and their correlation coefficient distribution intervals were calculated.

[0238] Determination of distribution characteristics:

[0239] Statistically analyze the correlation coefficients of all samples and observe their distribution characteristics (such as mean, standard deviation, and distribution range). Generally, the correlation coefficients of effective frequency-phase change rate pairs belonging to the same target are usually higher than a certain value (such as 0.7~0.8), while random combinations of different targets tend to be concentrated around 0.

[0240] Experience threshold setting and dynamic adjustment:

[0241] Based on statistical distribution, the judgment threshold is set at the dividing point between valid and invalid groups. The minimum value of positively correlated groups or the overall average value minus half a standard deviation is often used as an empirical threshold. For scenarios with high noise or low signal-to-noise ratio, a dynamic adjustment method can also be used to fine-tune the judgment threshold in real time according to the correlation distribution of the actual collected signals.

[0242] In a preferred embodiment of the present invention, for each frequency trajectory, its frequency change data and phase change data are respectively subjected to first-order difference to obtain a frequency change rate sequence and a phase change rate sequence, and correlation analysis is performed on the two at the same window position, including:

[0243] For each frequency trajectory, extract its frequency change data and phase change data within all time windows, and perform first-order difference processing according to the time window to obtain the corresponding frequency change rate sequence and phase change rate sequence.

[0244] Based on the frequency change rate sequence and the phase change rate sequence, at each time window position, the rate values ​​of the two are extracted respectively to construct a set of rate pairs under the same time window.

[0245] For rate pairs across all time windows, correlation analysis is performed using the correlation coefficient algorithm to obtain the correlation score sequence for each time window.

[0246] In this embodiment of the invention, for each frequency trajectory, the frequency change data and phase change data are respectively subjected to first-order difference to obtain the frequency change rate sequence and the phase change rate sequence, and the correlation analysis is performed on the two at the same window position, which can realize the multi-dimensional dynamic consistency detection of target motion characteristics and the quantitative elimination of abnormal trajectories.

[0247] The specific effects are as follows:

[0248] In underwater multi-target detection environments, the motion of different targets often exhibits a synchronous or coordinated trend of frequency and phase changes. By using first-order differential processing, subtle dynamics of signal change rates can be captured, enabling precise quantification of features such as acceleration, deceleration, and abrupt changes in moving targets.

[0249] Furthermore, within each time window, the rate of frequency change and the rate of phase change are paired to form a rate pair. A window sliding correlation algorithm is used to calculate the Pearson correlation coefficient or other correlation measure of the two sets of data window by window.

[0250] For example:

[0251] For physically consistent main frequency trajectories, the correlation coefficient will be significantly higher than the background noise in most windows (e.g., correlation > 0.8), reflecting the synchronicity of the changes between the two.

[0252] Abnormal trajectories (such as those caused by noise or false targets) are characterized by asynchronous frequency and phase changes or uncorrelated trends (such as correlation < 0.3), making them easy to remove directly.

[0253] Through the multi-window correlation quantization process described above, a synchronization index (such as the proportion of highly correlated windows) can be automatically calculated for each frequency trajectory, enabling effective differentiation between the true target's main trajectory and noise trajectories. Ultimately, this improves the overall accuracy and robustness of underwater moving target identification, significantly reduces false alarms and missed alarms, and enhances the algorithm's engineering applicability and intelligence level.

[0254] The above description represents the preferred embodiments of the present invention. It should be noted that, for those skilled in the art, various improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for analyzing underwater ultrasonic signals based on the Doppler effect, characterized in that, The method includes: The transmission frequency of the ultrasonic signal is obtained, and the ultrasonic signal is transmitted to the underwater target. The echo signals reflected back from multiple underwater targets are received, and echo signal data containing the reflection information of multiple underwater targets is obtained. The echo signal data is segmented according to a preset time window. Envelope demodulation and phase expansion operations are performed on the signal within each time window to extract its amplitude sequence, phase sequence and instantaneous frequency sequence data, and generate corresponding time series signal feature data. Based on the time series signal feature data, frequency change, amplitude change and phase change data are extracted, and clustering is performed to distinguish and label different underwater target components, resulting in a preliminary frequency shift signal set of multiple underwater targets. Each preliminary frequency shift signal set contains one or more frequency trajectories. Based on the preliminary set of frequency shift signals, and combined with amplitude and phase change data, the amplitude stability index and phase change synchronization index are calculated for each frequency trajectory in all time windows, and the main frequency trajectory is selected accordingly to form the final set of frequency shift signals. Based on the final frequency shift signal set, the frequency shift values ​​within each time window are extracted. The velocity of each underwater target within each time window is calculated based on the Doppler effect. The direction of motion is determined by combining the frequency change data and phase change data, and the motion state analysis results of the underwater targets are generated and output.

2. The underwater ultrasonic signal analysis method based on the Doppler effect according to claim 1, characterized in that, The echo signal data is segmented according to a preset time window. Envelope demodulation and phase unrolling operations are performed on the signal within each time window to extract its amplitude sequence, phase sequence, and instantaneous frequency sequence data, generating corresponding time-series signal feature data, including: The echo signal data is framed along the time axis, and the echo signal is divided into multiple frame segments according to a preset time window to obtain a segmented signal sequence. Based on the segmented signal sequence, envelope demodulation processing is performed on the echo signal of each frame segment to extract its amplitude variation curve and form amplitude sequence data. Based on the echo signal data, a phase unrolling operation is performed to extract continuous phase change curves, and the phase growth rate of each frame segment is calculated to obtain phase sequence and instantaneous frequency sequence data. The amplitude sequence, phase sequence, and instantaneous frequency sequence data are organized in chronological order to obtain the characteristic data of the time series signal.

3. The underwater ultrasonic signal analysis method based on the Doppler effect according to claim 1, characterized in that, Based on the time series signal characteristic data, frequency change, amplitude change, and phase change data are extracted, and clustering processing is performed to distinguish and label different underwater target components, resulting in a preliminary set of frequency shift signals for multiple underwater targets, including: Based on the characteristic data of the time series signal, the instantaneous frequency change, amplitude change and phase change data in each time window are extracted sequentially to form frequency change sequence, amplitude change sequence and phase change sequence respectively; Based on the frequency change sequence, the frequency change amount of adjacent time windows is analyzed, and points with similar change trends are grouped into preliminary frequency change groups to obtain multiple frequency change candidate groups. For each frequency change candidate group, its amplitude change sequence is extracted, and its mean and standard deviation are calculated respectively. Based on the condition that the mean amplitude is higher than the preset effective signal judgment threshold and the standard deviation is lower than the preset amplitude stability judgment threshold, the frequency change candidate groups that meet the amplitude criteria are selected. For each candidate group of frequency changes that meets the criteria, analyze its phase change sequence, calculate the correlation between the phase change trend and the frequency change trend within the group, and determine the group with a correlation higher than the preset correlation threshold as the preliminary frequency shift signal set of the same underwater target. Each set contains one or more frequency trajectories.

4. The underwater ultrasonic signal analysis method based on the Doppler effect according to claim 1, characterized in that, Based on the preliminary set of frequency-shifted signals, and combined with amplitude and phase change data, the amplitude stability index and phase change synchronization index are calculated for each frequency trajectory within all time windows. The dominant frequency trajectories are then selected based on these indices to form the final set of frequency-shifted signals, including: For each frequency trajectory, extract its amplitude variation data over all time windows and calculate its standard deviation as an amplitude stability index. For each frequency trajectory, based on the frequency change data and phase change data within all time windows, the frequency change direction and phase change direction within each time window are compared to determine whether the signs are consistent. The percentage of windows with consistent signs within all time windows is counted to obtain the consistency score of the growth direction. For each frequency trajectory, the frequency change data and phase change data are first-order differencing to obtain the frequency change rate sequence and the phase change rate sequence. Correlation analysis is performed on the two at the same window position, and the proportion of windows with correlation higher than the preset synchronization judgment threshold is used as the phase change synchronization index. For all frequency trajectories, combining the amplitude stability index, the growth direction consistency score, and the phase change synchronization index, a weighted scoring or threshold screening method is used to select the frequency trajectory with the highest comprehensive score as the main frequency trajectory. All main frequency trajectories are then collected to form the final frequency shift signal set.

5. The underwater ultrasonic signal analysis method based on the Doppler effect according to claim 1, characterized in that, Based on the final frequency shift signal set, the frequency shift values ​​within each time window are extracted. The velocity of each underwater target within each time window is calculated based on the Doppler effect. The direction of motion is determined by combining frequency change data and phase change data. The motion state analysis results of the underwater targets are generated and output, including: Based on the main frequency trajectory, the instantaneous frequency values ​​within each time window are extracted in the order of the time windows. Using the transmission frequency as a reference, the instantaneous frequency values ​​within each time window are differentially compared with the transmission frequency to obtain the frequency shift value of each time window. Based on the frequency shift values ​​within each time window of each main frequency trajectory, the Doppler effect calculation method is used to divide the frequency shift values ​​by the transmission frequency and multiply them by the propagation speed of ultrasound in water to obtain the corresponding radial velocity sequence within each time window. For each main frequency trajectory, based on the frequency change data and phase change data within each time window, determine the sign of the frequency change direction and the phase change direction within the same window. If they are consistent, determine that the underwater target is close to the source; if they are opposite, determine that the underwater target is far from the source. Generate motion direction criteria for each time window. The radial velocity sequence of each main frequency trajectory within all time windows is combined with the motion direction criterion to form the velocity and motion direction data of each underwater target over the entire time series. Based on the velocity and direction of motion data of each underwater target over the entire time series, the motion state analysis results of the underwater targets are generated and output.

6. The underwater ultrasonic signal analysis method based on the Doppler effect according to claim 2, characterized in that, Based on the segmented signal sequence, envelope demodulation processing is performed on the echo signal of each frame segment to extract its amplitude variation curve, forming amplitude sequence data, including: For each frame of the segmented signal sequence, for each sampling point, the absolute value of the signal is calculated with that of multiple adjacent sampling points, and the average of the calculated absolute values ​​of the signal is performed in the time direction to obtain the envelope amplitude corresponding to that sampling point. The envelope amplitudes of all sampling points in each frame are arranged sequentially according to the sampling time to form an amplitude variation curve. Based on the amplitude variation curve, the amplitude at each sampling time is extracted point by point to form amplitude sequence data.

7. The underwater ultrasonic signal analysis method based on the Doppler effect according to claim 2, characterized in that, Based on the echo signal data, a phase unrolling operation is performed to extract continuous phase change curves, and the phase growth rate of each frame segment is calculated to obtain phase sequence data and instantaneous frequency sequence data, including: Based on the echo signal data, for each sampling point, the preliminary phase value of the sampling point is calculated using its sine and cosine components, resulting in a set of preliminary phase sequence data; The initial phase sequence data is subjected to continuity correction. When the phase difference between adjacent sampling points is found to be greater than half a cycle, the abrupt change is eliminated by adding or subtracting 2π, so that the phase change curve remains monotonically continuous in time, forming continuous phase sequence data. Based on continuous phase sequence data, calculate the phase difference between every two adjacent sampling points, and divide the phase difference by the time interval between the two sampling points to obtain the phase growth rate sequence of each sampling point; Arrange the phase growth rates of all sampling points in chronological order of sampling time to form instantaneous frequency sequence data.

8. The underwater ultrasonic signal analysis method based on the Doppler effect according to claim 3, characterized in that, Based on the frequency change sequence, the frequency changes in adjacent time windows are analyzed. Points with similar trends are grouped into preliminary frequency change groups, resulting in multiple candidate frequency change groups, including: Based on the frequency change sequence, the frequency change within each two adjacent time windows is differentially calculated to obtain the frequency change increment corresponding to each time window. For the frequency change increments of all time windows, according to the principle of numerical similarity and consistent trend, the frequency change increments are clustered using the similarity criterion to obtain multiple frequency change subsequences with similar trends. Connect the frequency changes of adjacent time windows in each frequency change subsequence sequentially to form preliminary frequency change groups; Based on all the initial frequency change groupings, multiple frequency change candidate groups are generated.

9. The underwater ultrasonic signal analysis method based on the Doppler effect according to claim 3, characterized in that, For each candidate group of frequency changes that meets the criteria, its phase change sequence is analyzed, and the correlation between the phase change trend and the frequency change trend within the group is calculated. Groups with a correlation higher than a preset correlation threshold are identified as the initial set of frequency shift signals for the same underwater target, including: For each candidate group of frequency changes that meets the criteria, extract the phase change sequence within its corresponding time window to form the phase change data of that group. For the frequency change sequence and phase change sequence of the candidate frequency change group, the first-order difference method is used to obtain the frequency change rate sequence and phase change rate sequence, respectively. Correlation analysis was performed on the frequency change rate sequence and the phase change rate sequence within each corresponding time window, and the correlation coefficient was calculated to obtain the correlation criterion for each group. For each frequency change candidate group, determine whether its correlation criterion is higher than the preset correlation threshold. Groups that are higher than the threshold are determined as the preliminary frequency shift signal set of the same underwater target, and output the preliminary frequency shift signal set that meets the conditions.

10. The underwater ultrasonic signal analysis method based on the Doppler effect according to claim 4, characterized in that, For each frequency trajectory, first-order differences are performed on its frequency variation data and phase variation data to obtain frequency variation rate sequences and phase variation rate sequences, respectively. Correlation analysis is then conducted on the two sequences at the same window location, including: For each frequency trajectory, extract its frequency change data and phase change data within all time windows, and perform first-order difference processing according to the time window to obtain the corresponding frequency change rate sequence and phase change rate sequence. Based on the frequency change rate sequence and the phase change rate sequence, at each time window position, the rate values ​​of the two are extracted respectively to construct a set of rate pairs under the same time window. For rate pairs across all time windows, correlation analysis is performed using the correlation coefficient algorithm to obtain the correlation score sequence for each time window.

Images